Multiple unmanned aerial vehicle coordinated strikes against ground targets based on an improved multi-agent deep deterministic policy gradient algorithm
Saved in:
| Title: | Multiple unmanned aerial vehicle coordinated strikes against ground targets based on an improved multi-agent deep deterministic policy gradient algorithm |
|---|---|
| Authors: | Wei Li, Xin Chen, Wei Yu, Mingyang Xie |
| Source: | Proceedings of the Institution of Mechanical Engineers, Part I: Journal of Systems and Control Engineering. |
| Publisher Information: | SAGE Publications, 2024. |
| Publication Year: | 2024 |
| Subject Terms: | 0203 mechanical engineering, 0103 physical sciences, 02 engineering and technology, 01 natural sciences |
| Description: | With the development of swarm intelligence, multi-unmanned aerial vehicle (UAV) cooperation in uncertain dynamic environments has been able to safely and efficiently realize all-round strikes on ground targets. However, the limited life span of UAV onboard batteries challenges the cooperative strikes against the ground targets. To address this challenge, this paper proposes an improved multi-agent deep deterministic policy gradient (i-MADDPG) algorithm for energy-saving cooperative strikes of ground targets. Introducing a sensor fusion layer and self-attention mechanism to the actor network helps the UAVs collect more comprehensive information and filter the collected data, resulting in more accurate environmental perception. In addition, importing an egocentric state representation mechanism into the critic network contributes to the computation of different Q-values for different UAV observation states, which ensures each UAV can make appropriate decisions based on its own state. Simulation experiments are conducted to validate the performance of the improved MADDPG algorithm. The results show that the success rate of ground target strikes of the proposed algorithm is improved by 29.5%, and the collision rate is reduced compared with the traditional MADDPG algorithm. |
| Document Type: | Article |
| Language: | English |
| ISSN: | 2041-3041 0959-6518 |
| DOI: | 10.1177/09596518241291185 |
| Rights: | URL: https://journals.sagepub.com/page/policies/text-and-data-mining-license |
| Accession Number: | edsair.doi...........1756f1be2ce6c4be48ce0305f86f2c3d |
| Database: | OpenAIRE |
Nájsť tento článok vo Web of Science | Abstract: | With the development of swarm intelligence, multi-unmanned aerial vehicle (UAV) cooperation in uncertain dynamic environments has been able to safely and efficiently realize all-round strikes on ground targets. However, the limited life span of UAV onboard batteries challenges the cooperative strikes against the ground targets. To address this challenge, this paper proposes an improved multi-agent deep deterministic policy gradient (i-MADDPG) algorithm for energy-saving cooperative strikes of ground targets. Introducing a sensor fusion layer and self-attention mechanism to the actor network helps the UAVs collect more comprehensive information and filter the collected data, resulting in more accurate environmental perception. In addition, importing an egocentric state representation mechanism into the critic network contributes to the computation of different Q-values for different UAV observation states, which ensures each UAV can make appropriate decisions based on its own state. Simulation experiments are conducted to validate the performance of the improved MADDPG algorithm. The results show that the success rate of ground target strikes of the proposed algorithm is improved by 29.5%, and the collision rate is reduced compared with the traditional MADDPG algorithm. |
|---|---|
| ISSN: | 20413041 09596518 |
| DOI: | 10.1177/09596518241291185 |